Readme
======

The file Replication_PiercingThroughOpacity.do contains the regression codes used to construct all tables and figures in "Piercing through Opacity: Relationships and Credit Card Lending to Consumers and Small Businesses during Normal Times and the COVID-19 Crisis."

The supervisory confidential Federal Reserve Y-14M data used in the regressions cannot be shared publicly to protect the confidentiality of the banks that submitted their information to the Federal Reserve for the purposes of stress tests. Even within the Federal Reserve, only those with an approved data need can see Y-14 data and this dataset. 

All regressions in the paper are run using STATA for random samples drawn from the Y-14M dataset and aim to assess how relationships affect credit card terms to credit card customers during normal times (June 2013 to February 2020) and the COVID-19 times (November 2019 to June 2020). 

Please refer to Section 3 in the paper for our data construction and sample selection. 

Our main analyses use a 0.5% random sample for new consumer originations and a 5% random sample for new small business originations.

Some analyses in the Online Appendix use a 0.1% random sample for existing consumer accounts and a 1% random sample for small business existing accounts.

Table 1 and Table OA.4 contain definitions and data sources for all variables used in the analysis.

All codes are anonymized to ensure that we do not expose information on any individual banks, loans, or customers, or Federal Reserve file paths.


